Do Machine Learning Models Represent Their Targets?

Abstract

I argue that ML models used in science function as highly idealized toy models. If we treat ML models as a type of highly idealized toy model, then we can deploy standard representational and epistemic strategies from the toy model literature to explain why ML models can still provide epistemic success despite their lack of similarity to their targets.

Reliability in Machine Learning

“…8 There is also a rich debate regarding the nature of ML opacity itself (Boge, 2022;Creel, 2020), but for this we direct our readers to Buchholz, 2023). 9 Though Sullivan (2023) argues that ML models can be assessed in a similar way as simulations, or 'toy models.' 10 There is a rich debate on whether there is actually a difference between opacity in human reasoning and the reasoning of machine learning models and whether this difference matters.…”

“…9 9 Though Sullivan (2023) argues that ML models can be assessed in a similar way as simulations, or ‘toy models.’ …”